Association of CT Findings with clinical severity in Patients with COVID-19, a multicenter Cohort observational study

Background 2019 Novel Coronavirus disease (COVID-19) may cause critical illness including severe pneumonia and acute respiratory distress syndrome. Our purpose is to was to analyze the radiological features of COVID-19 pneumonia and its association with clinical severity. Methods This retrospective study included 212 patients (122 males, Mean age, 45.6 ± 12.8 years) from 10 hospitals. Chest CT, chest X-ray (CXR), clinical and laboratory data at admission and follow-up CT were collected. Chest CT and CXR were reviewed and CT score of the involved lung was calculated. Results 94.3% patients had pneumonia on the baseline CT at admission. The most CT findings were as follows: GGO (140/200), GGO with consolidation (38/200) and consolidation (16/200) most involving the lower lobes with a predilection for the peripheral aspects. The CT score negatively correlated with Lymphocyte count while it positively correlated with C-reactive protein. ROC curve showed an optimal cutoff value of the CT score of 15 had a sensitivity of 70% and a specificity of 96.5% for the prediction of severe status. Series CT showed GGO or consolidation gradually reduced in 52 patients while 6 patients had reticular opacities. 14 patients showed the normal CXR while GGO were found on CT. insufficient for evaluation


Introduction
In December of 2019, there began an epidemic of acute febrile respiratory illness and caused by a new coronal virus in China. The first cases were reported to be occurred in Huanan Seafood Market of Wuhan, Hubei province, China, and the infection rapidly spread nationwide during Chinese Spring Festival (1,2). The World Health Organization (WHO) named the disease as 2019 Novel Coronavirus disease (COVID-19)(3),which was caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (4) and it shares 79.5% sequence identity with SARS-CoV (5). After series of measures including Wuhan City lockdown, the quarantine and isolation, the number of cases in China has been significantly reduced. Although in most cases, the symptoms of COVID-19 are mild and run a selflimiting course; however, in a small proportion of individuals,the illness progressed rapidly to severe pneumonia with acute respiratory distress syndrome which can result in respiratory failure and death.
By April 4, 2020, more than one million cases of COVID-19 pneumonia had been reported worldwide, and 58901 patients had died, equivalent to a mortality rate of approximately 5.3%. (6).
The diagnosis of viral infection is based on identification of the virus, however, the recognition of some imaging features of the disease can become useful,especially to evaluation of the lung involvement. The CT features of COVID-19 pneumonia in Wu Han were successively reported (7)(8)(9)(10)(11).
But cases identified in other Chinese cities have not been described moreover, these study only reports chest CT findings of COVID-19 pneumonia. The aim of this study, therefore, was to study the correlation of CT findings of COVID-19 pneumonia with clinical, laboratory findings, and then to compare chest X-Ray (CXR) and CT in detection of COVID-19 pneumonia.

Study population
This study was approved the institutional review boards of our hospitals (2020-21-K16). The institutional review boards waived written informed consents for this retrospective multicenter study. This retrospective study included 212 patients with COVID-19 from 10 hospitals of six provinces between January and February 2020. SARS-CoV-2 infection was confirmed by positive test for viral RNA in aspiration material or nasopharyngeal or oropharyngeal swab specimens or bronchoalveolar lavage fluid collected from patients by next-generation sequencing or standard real-time reverse transcription polymerase chain reaction protocol. Baseline laboratory Findings including white blood cell count (WBC),Neutrophil count Lymphocyte count and C-reactive protein (CRP) and the CT scan at admission were collected. According to Diagnosis  Healthcare; Toshiba Aquilion ONE TSX-301C/320, Tokyo, Japan; Optima CT680 Series, GE MEDICAL SYSTEMS, America;) by using a standard CT protocol. The whole chest was craniocaudally scanned from lung apex to the lowest hemidiaphragm during a single breath-hold. Scan parameters were as follows: Scan parameters were as follows: The following protocols were used: 120 kVp, reference tube current of 110-250 mAs, detector collimation of 16-320 × 0.5-0.625 mm, slice thickness of 1.5-2 mm, and slice interval of 1.5-2 mm, pitch of 1-1.375, a high spatial resolution reconstruction algorithm.

CT Images Interpretation
All CT images were evaluated by two thoracic radiologists (M.L and H.Z with 15 years and 10 years experiences in chest imaging) in consensus. In case of disagreement, the third senior radiologist (Y.G) with 37-years experiences in the chest radiology) made the final decision.
CT findings were evaluated for the type of radiologic pattern(13) :①parenchymal attenuation disturbances;②Ground-glass opacity(GGO): defined as slightly increased attenuation of the lung parenchyma that is unrelated to the obscuration of the vessels and adjacent airway walls. ③Consolidation: defined as increased attenuation of the lung parenchyma, resulting in the obscuration of the vascular outlines and adjacent airway walls; ④Nodules and tree-in-bud opacities; ⑤Crazy-paving pattern: defined as interlobular septal thickening superimposed on ground-glass opacities; ⑥Reticulation: defined as thin linear opacities, which correspond to the thickened peripheral connective septa. Distribution: unilateral/bilateral, upper/middle/lower, and central/peripheral/ peribronchovascular/random. Extent: focal (n ≤ 2)/multifocal(n ≥ 3) /diffuse(continuous involvement of at least 2 lobes); Involved number of lung segments. In addition, the presence of pleural or pericardial effusion, mediastinal lymphadenopathy defined as a lymph node ≥ 1 cm in short-axis diameter) were evaluated.
In order to quantitative analysis the extent of lung involvement with CT score, we modified the CT score of opacity proposed by Wu et al (14) based on the size and intensity of opacity: size score-if more than half of the segment on the biggest scope of lesion level on axial section CT was involved, then the segment (2 point) was recorded as being involved. If less than half of the segment on the biggest scope of lesion level on axial section CT was involved, then the segment (1 point) was recorded as being involved; intensity score-if the lesion is GGO or crazy-paving sign, the intensity (1 point) was recorded. If lesion is consolidation, the intensity (2 point) was recorded. Severity Score of each lesion = lesion size score × lesion density score. Summation of scores provided overall lung involvement on CT. Maximal CT score for both lungs was = 2(consolidation)x 2 size x20 segments  Figure 1 showed abnormal opacity on the baseline CT were found in 30 patients without fever and 14 asymptomatic patients. Table 1 demonstrated the frequency of the baseline chest CT findings. The most CT findings were as follows (Fig. 2): focal or multifocal GGO (140/200) most commonly involve the lower lobes and show a predilection for the peripheral aspects of the lungs. GGO with patchy consolidation (38/200) was the other significant sign. Crazy-paving sign was found in 6 patients, however, no reticulation, nodules or tree-in-bud opacities were found at admission.  In this cohort, the most common clinical symptom are fever, dry cough which were consistent with our previous reports (15). Lymphocytopenia and increase of CRP were reported to be the most common laboratory findings (16, 17). The normal white blood cell and lymphocyte count and in Korea (19), they found chest radiography was negative but CT scans performed on the same day showed bilateral, multiple GGOs. And Three of the nine patients (33.3%) had parenchymal abnormalities detected by chest radiography were reported by his team (20). In our study, there were no abnormalities on the initial CXR at admission in 14 patients while their baseline CT showed focal or multifocal GGO. And CXR in 1 patient could not be evaluated due to old tuberculosis. Thus,we believe that the high-risk population can benefit from the early CT scan and CXR is yet insufficient for evaluation of the suspected COVID-19 pneumonia.
Chung et al (8) showed three patients (21%, 3/21) with normal CT scans, 12 (57%) with GGO only, and 6 (29%) with GGO and consolidation in appearance, 15 (71%) patients with two or more lobes involved, and 16 (76%) with bilateral disease. Song et al (21) found pure GGO in 77% patients, GGOs with consolidation in 59% cases, bilateral GGOs in 88% of patients. Similarly, we showed that the most common CT findings of COVID-19 pneumonia in the early stage were focal (36%) or multifocal peripheral GGO (51%) with/without patchy consolidation in a lower lung involvement, however, crazypaving sign and reticulation were not common. Although some of these findings were similar with SARS-CoV, MERS, H1N1 pneumonia (22)(23)(24), cautious attention to high exposure risk and CT findings are helpful for early suspicion of COVID-19 pneumonia and isolation as soon as possible.
CT or radiologic scores of involved lung was reported to be associated with clinical outcomes (14,(25)(26). Our modified CT score takes into account the number, size and intensity of lesions, based on the previous method and we found the modified CT score negatively correlated with Lymphocyte count while it positively correlated with C-reactive protein, moreover, the baseline CT scores were higher in severe group than in non-severe group. Regarding the prediction of clinical severity, we found an optimal cutoff value of a CT score of 15 with sensitivity of 70% and specificity of 96.5%.
Further study was needed to demonstrate whether this modified CT score predict the lung function and clinical outcomes.
The serial CT scans provided an opportunity to observe the longitudinal lung changes of COVID-19 pneumonia. According to the follow-up CT images, the median time interval to gradually resolved opacity in 52 patients was 8 days. Although the third CT scan of 6 patients showed the features suggestive of lung fibrosis with reticular opacities, architectural distortion and bronchial dilatation.
There are some limitations in our research. First, these patients came from multicenter outside Hubei province, thus they cannot represent real condition in whole country. There are remarkable selection bias because most patients were common type, thus, our study could not comprehensively reflect the difference of clinical and CT findings of patients in mild, common, severe and critical type. Another significant limitation is that, only some of our patients underwent a series of CT scans, there were significant differences in the time interval of CT follow-up and treatment options in each center.
Therefore, the characteristics of temporal CT changes and the treatment effect on CT findings cannot be clarified in detail.